lucky13
Group: Members
Posts: 1478
Joined: Feb. 2007 |
 |
Posted: July 28 2007,15:17 |
 |
| Quote | | He had tons of user reports, I felt... |
That's what the developers don't want. They want benchmarks that are quantitative, unbiased, objective, etc. Not feelings, not user reports, not "this happened here, but not here." Some criteria that can be presented that shows in black and white that one option is better than another.
http://www.onegoodmove.org/fallacy/causal.htm
By way of analogy (only regarding anecdote and not the merits of proposed codes), I can say that one brand of soap doesn't dry out my skin as much as another brand. How do I know this? Maybe I can "feel" some difference. But when I look at the ingredients for both brands, they're virtually identical. So I can start looking at other factors. I tried Brand A soap in winter when the air wasn't as humid as it is now, and now I'm using Brand B (100% humidity today). I feel a difference. Is it the soap or is it the climate?
The same kind of fallacy is true of snakeoil. Take some magical potion or herb (e.g., echinacea) and you won't get sick. Folks swear by it, so it must be true -- at least until there are studies done to see if it really works, and it usually doesn't.
From everything I've read, the empirical evidence was lacking about whether Kolivas' patch had any measurable benefit over any other scheduler. The developers wanted more than "I clicked on this and it stuttered, but I applied the patch and now it doesn't." That's why his scheduler code wasn't (yet) mainlined.
See this also:
http://kerneltrap.org/node/14008
I strongly disagree with Kolivas' position that desktops are abandoned by 2.6+ development trends. This is like saying that Microsoft abandoned desktops by releasing Server 2008 and CE. With either OS, you have basically the same kernel and code base that can be configured in various ways to suit particular needs. These are, for lack of better ways of explaining, one-size-fits-all. They're just custom-tailored for servers, desktops, and mobile devices. With respect to Linux, that goes beyond distro-compiled kernels and includes custom patches like -ck, grsecurity, etc.
Excluding Kolivas' code from mainline does NOT mean Linux is taking a server-only approach -- it means Linus and other developers aren't convinced (yet) that Kolivas' approach is necessarily the best or even ready for inclusion. Everything I've read showed they had, for the most part, open minds and were continually willing to consider and review what he was suggesting.
It's now irrelevant if he's really walking away from even developing his patches and continuing to offer them as he has unless someone else picks up where he left off. Some of his ideas will probably be implemented eventually. It just won't be from his efforts.
| Quote | | Iso became 93mb, I removed lots of stuff but also added gtk2, firefox2, java, flash 9.. |
I had a hunch it would be in the ~100MB (or more) range.
--------------
"It felt kind of like having a pitbull terrier on my rear end."
-- meo (copyright(c)2008, all rights reserved)
|
.
DSL Market ,
Great VPS hosting provided by Tektonic
